Binary Quality Calibration
Instead of documenting what 'quality' means, classify shipped work as high-quality or not — and let examples teach the bar.
- Difficulty
- Easy
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 90%
Miro spent a year writing long documents trying to define 'high quality' and got stuck in endless conversations. The breakthrough was to stop describing and start classifying: every month the design leadership team triages everything shipped into a binary bucket — high quality or not — with the reasons attached, and shares examples. Like teaching colors by showing pink and red rather than describing them, the shown examples calibrate a shared, intuitive sense of quality across the team.
Origin
Developed at Miro under Varun Parmar after a heavy, document-driven attempt to define quality failed. Parmar draws the analogy to how modern AI systems learn via classification.
Core principles
- 01Some things (like quality) are far easier to show with examples than to describe in prose.
- 02A binary label plus a reason beats a long rubric for building shared judgment.
- 03Calibration is a muscle built through repeated, concrete examples — not a one-time definition.
- 04Start lightweight and operationalize later rather than front-loading a heavy framework.
How to run it
- 1
Stop trying to fully define the abstract standard
Abandon the attempt to exhaustively document quality attributes and how to measure them — it produces endless conversations and a heavy process.
Watch out Writing ever-longer definition documents feels productive but can stall the org for months.
- 2
Assign a single owner and a regular cadence
Designate one design leader responsible for making the triage happen on a recurring (e.g. monthly) basis across everything shipped.
- 3
Classify each shipped item as high-quality or not
Apply a simple binary function to each shipped piece, and for the 'not high quality' cases, record the specific reasons (a, b, c, d, e).
Pro tip Keep it binary — the forcing function is the simplicity, not a nuanced score.
- 4
Share examples to calibrate and democratize
Make the classified examples available to other designers so they build pattern-matching over time, then progressively democratize access across the org.
Pro tip Show three examples of 'pink' and three of 'red' — concrete pairs teach the distinction words can't.
In the wild
After a year of trying to write down what quality meant and getting into endless conversations, Miro switched to classifying shipped work as great vs not great. Parmar compares it to describing colors: you can't explain pink versus red in words, but shown three examples of each, anyone instantly gets it.
→ A lightweight approach that worked 'decently well' to align design leadership on what high quality means, replacing a stalled heavy documentation effort.
Common mistakes
Solving quality with ever-longer definition documents
Miro had leaders write long documents defining quality attributes and measures; it felt like codifying but was a very heavy way to solve the problem and left endless disagreement. Showing examples resolved it faster.
Skipping operationalization after the pilot
Parmar notes the approach still needs to be operationalized and democratized so everyone has access; a binary triage that stays siloed in design leadership doesn't build org-wide judgment.
Is it for you?
Best for
Design and product teams struggling to align on a subjective quality bar that resists written definition.
Not ideal for
Quality dimensions that are genuinely objective and measurable (performance budgets, error rates), where metrics beat binary human classification.
From the transcript
“our design leadership team does a triage of everything that got shipped into like high quality or not high quality it's just like a binary…”
“if I show you and I say Lenny these are three examples of what pink is and these are three examples of red is then…”
“we had like a bunch of our leaders go and like write like you know documents like really long documents in terms of like what…”
From the episode
An inside look at how Miro builds product: Lessons on outmaneuvering competitors, team structure, product quality, and moving fast
Varun Parmar (CPO of Miro)